Screening of Immune-relate Genes and Prognosis of Lung Adenocarcinoma based on Ensemble Learning

Lung adenocarcinoma (LUAD) is one of the most significant malignant tumors threatening human health. Research has shown that certain immune-related genes (IRGs) play a crucial regulatory role in the onset and progression of LUAD. However, accurately identifying and utilizing these genes for prognosis prediction remains challenging. In this study, we developed a Multivariate Fusion IRGs Screening model (MFIRGS) to investigate the impact of IRGs on the survival time of LUAD patients.We employed multivariate CoxPH regression analysis to reduce gene data dimensionality and eliminate redundant information interference. XGBoost nonlinear function fitting was subsequently used to evaluate the influence of IRGs on LUAD occurrence and development, allowing for precise identification of key IRGs. LASSO regression analysis was applied to further exclude IRGs weakly associated with LUAD prognosis. Based on these results, we designed a prognostic model (LightGBM) incorporating the Histogram Algorithm (HA), Exclusive Feature Bundling (EFB), Gradient-based One-Sided Sampling (GOSS), and a Leaf-wise growth strategy. Patients were categorized into three groups using the K-means algorithm based on IRG expression levels, achieving 5-year survival rate classifications of over 65%, between 30% and 65%, and below 30%.The IRGs identified through the MFIRGS model enabled the construction of an effective prognostic model for LUAD patients. This model offers more accurate 5-year survival rate predictions, providing valuable insights for clinical decision-making and enhancing prognosis accuracy for LUAD patients.